This post has been republished via RSS; it originally appeared at: New blog articles in Microsoft Tech Community.
The Azure Machine Learning team is excited to announce that Microsoft partner Seeq has developed an open-source integration that integrates Seeq Workbench with Azure Machine Learning. The Seeq Workbench application is used by process manufacturers (those who produce consumer goods from raw materials following a specific process such as Food and beverage, Oil & Gas, Pharmaceuticals, etc.) to understand their data, analyze patterns and trends, monitor assets and interact with their data in real time. This Azure Machine Learning - Seeq integration allows to deploy machine learning models directly into Seeq Workbench, putting their models in the hands of engineers and process experts faster. Engineers and process experts can incorporate the results into their analysis and decision-making workflow, informing and generating insights to improve production processes, achieve sustainability targets and drive key business decisions.
“The Seeq Azure Machine Learning integration bridges the gap between IT and OT teams – placing the value of Azure ML inferences directly in the hands of process engineers in manufacturing organizations to accelerate productivity and streamline operations.” - Brian Parsonnet, CTO, Seeq Corporation
Streamlining Operationalization of Azure ML Models
Many data scientists are comfortable with Azure Machine Learning for machine learning operations and model lifecycle management but are relatively unfamiliar with manufacturing-specific data and analytics techniques. Engineers and process experts, on the other hand, understand the data but are not familiar with data science techniques or tools. This divide results in models that are either inaccurate or not trusted by the people that need to act based on the insights generated.
This new integration brings these two worlds together and lets the data scientists create their models in Azure Machine Learning, deploy the models through Managed Online Endpoints, and then publish those models into the Seeq Workbench with the click of a button. Seeq makes it easy for engineers and process experts to evaluate and provide the necessary context and feedback for these models. This fosters collaboration between the two sides, builds rapport and results in solutions that are relevant and effective for the business.
“Seeq and Azure Machine Learning are critical and complementary solutions for a successful machine learning model lifecycle,” says Megan Buntain, Director of Cloud Partnerships at Seeq. “By capitalizing on IT and OT users’ strengths, the Seeq Azure Add-on expands the Seeq experience and creates new opportunities for organizations to scale up model deployment and development.”
Figure 1: Azure Machine Learning and Seeq toolbench working together
What’s available now
Data scientists can now build and train their model in Azure Machine Learning utilizing all the tools and capabilities required to build robust, production grade models using their preferred authoring experience and frameworks. Once they are ready to get feedback or push the model to production use the data scientist will deploy their model via the new managed online endpoints feature in Azure Machine Learning.
Engineers and process experts using Seeq Workbench can leverage the Azure Machine Learning Add-on and access the models the data science teams have deployed. This enables them to:
- Annotate their data sets and results with comments, take snapshots of the data and share those comments with the data scientists building the models to help improve and refine the model output.
- Embed the models into existing workflows within Seeq and run inferencing over their selected data either on an adhoc basis or on a recurring schedule.
- Publish the results of the model into existing dashboards for decision makers to have a single pane of glass view of the data and insights.
Consider a customer who wants to create a predictive maintenance model for a critical asset. An engineer/process expert can provide required context using Seeq Workbench i.e., what signals to include, what time periods to use for model training, what time periods to avoid, etc. A data scientist can import the relevant data into Azure Machine Learning, train a predictive maintenance model using the data set identified by the engineer/process expert, and deploy it via a managed online endpoint. The engineer/process expert can access created models using the Add-on and evaluate performance in Seeq Workbench typically using historical time periods of interest. Findings and feedback can be captured in Seeq Workbench for review and refinement by the data scientist. This can be an iterative process. Once the engineer/process expert is satisfied with the performance of the model, they can operationalize the model from the Add-on to monitor the asset in near real time.
Figure 2: Architecture for Azure ML and Seeq Integration
The partnership between Microsoft and Seeq results in a solution for end-users that::
- Streamlines operational data access via advanced analytics and ML.
- Provides self-service operational analytics and ML for engineers and process experts
- Improves collaboration between teams.
- Enhances monitoring, diagnostics, and root cause analysis.
- Enables real-time decisions that drive business outcomes.
- Provides quicker time to value for OT and IoT data analytics
For more information on Azure ML and Managed Online Endpoints see the links below:
- Deploy an ML model by using an online endpoint (preview) - Azure Machine Learning | Microsoft Docs
- Use managed online endpoints (preview) in the studio - Azure Machine Learning | Microsoft Docs
- Announcing managed endpoints in Azure Machine Learning for simplified model deployment - Microsoft Tech Community
For more information on the Seeq integration:
- GitHub - seeq12/seeq-azureml: Seeq - Azure ML integration example
- User Guide — seeq-azureml 0.1.0 documentation (seeq12.github.io)